Human beings have a weird, poorly-understood ability to pick a single conversation out of a noisy room, it's called the "cocktail party effect" and while its exact mechanism isn't totally understood, researchers do know that vision plays a role in it, and that being able to see the speaker helps you pick their words out of a crowd. Read the rest
The Echo Look is the next version of the Alexa appliance: it has an camera hooked up to a computer vision system, along with its always-on mic, and the first application for it is to watch you as you dress and give you fashion advice (that is, recommend clothes you can order from Amazon). Read the rest
Nearly 250 million people in the world have impaired vision. Oxsight is developing augmented reality glasses that could supplement or even replace canes and seeing eye dogs for many. Read the rest
Dan Howland writes, "Screengrabs that made me laugh from Google (Happiest Place on) Earth. It does pretty well with conventional architecture, but it freaked out on the Haunted Mansion's cupola and chimblies, and Dumbo just looks like a Jello salad."
Adam Harvey, creator of 2012's CV Dazzle project to systematically confound facial recognition software with makeup and hairstyles, presented his latest dazzle iteration, Hyperface, at the Chaos Communications Congress in Hamburg last month. Read the rest
Weird twitterbot herder Shardcore writes, "@everytrumpette draws from the large corpus of photographs of the attendees of Trump rallies. A face detection algorithm identifies a member of the crowd, and then zooms in. Who are these people? How can they not only accept, but openly embrace an ideology of hate? This bot examines them, one by one, to try and see the humanity." Read the rest
Yaskawa celebrated its 100th birthday by pitting one of its Motoman-MH24s against a samurai in a sword-swinging contest, proving once and for all that if you need to chop bamboo into small pieces without having to walk around, you need a robot, not a human. Read the rest
Sterling Crispin uses evolutionary algorithms to produce masks that satisfy facial recognition algorithms: "my goal is to show the machine what it’s looking for, to hold a mirror up to the all-seeing eye of the digital-panopticon we live in and let it stare back into its own mind." Read the rest
In Xerox scanners/photocopiers randomly alter numbers in scanned documents, computer scientist David Kriesel shows that the Xerox WorkCentre 7535 randomly changes the numbers in its scans. The copier has firmware that tries to compress images by recognizing the numbers and letters in the documents it scans, and when it misinterprets those numbers, it produces untrustworthy output. The bug also occurs in the Xerox 7556 and possibly other machines, and as Kriesel points out, this could mean that engineering diagrams, invoices, prescriptions, architectural drawings and other documents whose numeric values are potentially a matter of life-and-death (or at least financial stability) are being randomly edited by machines we count on to produce faithful copies. Read the rest
Fast, Accurate Detection of 100,000 Object Classes on a Single Machine a prizewinning paper by Google Research scientists, describes a breakthrough in machine vision that can distinguish between a huge class of objects 20,000 times faster than before.
This so-called convolution operator is one of the key operations used in computer vision and, more broadly, all of signal processing. Unfortunately, it is computationally expensive and hence researchers use it sparingly or employ exotic SIMD hardware like GPUs and FPGAs to mitigate the computational cost. We turn things on their head by showing how one can use fast table lookup — a method called hashing — to trade time for space, replacing the computationally-expensive inner loop of the convolution operator — a sequence of multiplications and additions — required for performing millions of convolutions with a single table lookup.
We demonstrate the advantages of our approach by scaling object detection from the current state of the art involving several hundred or at most a few thousand of object categories to 100,000 categories requiring what would amount to more than a million convolutions. Moreover, our demonstration was carried out on a single commodity computer requiring only a few seconds for each image. The basic technology is used in several pieces of Google infrastructure and can be applied to problems outside of computer vision such as auditory signal processing.
Fast, Accurate Detection of 100,000 Object Classes on a Single Machine (via /.)
(Image: Clutter, a Creative Commons Attribution Share-Alike (2.0) image from neofob's photostream) Read the rest
Martin Backes is selling a limited edition of 333 "Pixelhead" anonymity masks, which allow you to replace your face with the pixellated likeness of German Secretary of the Interior Hans-Peter Friedrich. Masks are made to order and to measure, take 4-6 weeks for delivery, and cost €158 with shipping.
The full face mask Pixelhead acts as media camouflage, completely shielding the head to ensure that your face is not recognizable on photographs taken in public places without securing permission. A simple piece of fabric creates a little piece of anonymity for the Internet age. The material used is elastic fabric for beach fashion and sports gear with a fashionable Pixel-style print of German Secretary of the Interior Hans-Peter Friedrich. The mask has holes for your eyes and mouth, so you can see and breathe comfortably while wearing the mask, secure in the knowledge that your image won’t be showing up anywhere you don’t want it to.
Timo's video "Robot readable world" is made up of stitched-together found footage from computer vision systems, "exploring the aesthetics of the robot eye." It was inspired by Matt Jones's essay The Robot-Readable World, and it reminds me Laura Mulvey's idea of the Male Gaze.
